Howdy, it’s Barsee.

Happy Thursday, AI family, and welcome to another AI Valley edition. This issue takes 6 minutes to read.

Today’s climb through the Valley reveals:

• Google cuts off 90% of the internet from AI

• Google releases Gemini 2.5 Computer Use

• Figure unveils next-gen humanoid robot

• Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

WISPR FLOW

The speed upgrade for every thought

Most people lose their best ideas between thinking and typing. Wispr Flow closes that gap. It turns your voice into clean, structured text that's instantly ready for Slack, Notion, Gmail, or ChatGPT (no editing, no delay).

You can dictate notes, rewrite messages, or even draft code without ever breaking focus. It's one of those tools that quietly becomes indispensable: the faster you talk, the faster you work.

Try Wispr Flow → Turn ideas into action instantly.

_{*This is sponsored}

THROUGH THE VALLEY

1/ Google cuts off 90% of the internet from AI

Last month, Google quietly removed the num=100 search parameter, the option that let users and crawlers view 100 results at once instead of the default 10. It sounds minor, but it’s not: AI systems from OpenAI, Anthropic, and Perplexity depend on Google’s index for retrieval and training. With this change, those systems can now see only a tenth of what they used to.

The fallout was fast: Search Engine Land reports 88% of websites lost visibility as rankings beyond the top 10 vanished from view. For startups, the shift makes organic discovery harder; even good products now struggle to be found unless they already rank near the top.

Why does it matter?

AI models and search retrieval systems depend on large-scale, diverse web data for training and serving answers. With this change, the diversity and breadth of training data is sharply reduced, increasing bias and weakening organic discovery of non-mainstream sites, niche communities, and emerging startups. Sites like Reddit that typically rank below the top results saw a noticeable drop in AI-attributed citations, underscoring the dependence of LLMs on the availability of deep web content.

2/ Google releases Gemini 2.5 Computer Use

Google DeepMind has launched Gemini 2.5 Computer Use, a powerful AI model that directly operates web and mobile UIs visually (clicking, scrolling, typing, filling forms, navigating dropdowns, and even functioning behind logins). This makes it possible for agents to complete real multi-step workflows, like transferring CRM data, organizing notes, and automatically testing user interfaces, all through a built-in computer-use tool available in the Gemini API (Google AI Studio) and Vertex AI.

Gemini 2.5 Computer Use runs in a continual loop: it receives a screenshot, a user request, and the history of previous actions, then returns a UI action (e.g., click, drag, type). After every action, the interface updates and sends a new screenshot to the model, repeating until the task is finished. This method enables robust step-by-step automation and dynamic adaptation to changing UI states.

Benchmarks show Gemini 2.5 Computer Use outperforms rivals such as Anthropic’s Claude Sonnet 4 and OpenAI agents. It achieves 65.7% accuracy on Online-Mind2Web, 79.9% on WebVoyager, and 69.7% on AndroidWorld. It also delivers lower latency (around 225 ms per step), which is crucial for production environments where speed matters.

Why does it matter?

The launch of Gemini 2.5 Computer Use is a significant milestone, signaling Google’s full-scale entry into the agentic AI race against OpenAI and Anthropic. By empowering models to visually control browsers and UIs, Google shifts agentic AI from passive conversation to true autonomous action.

PEAK OF THE DAY

Figure unveils next-gen humanoid robot

Figure just launched Figure 03, its most advanced humanoid robot yet, designed for homes, factories, and the world at scale. It’s a full rebuild, with new hardware, smarter software, and one big goal: move and think more like a human.

• The robot learns by watching and mimicking people. Helix, Figure’s custom vision-language-action AI, directly enables rapid adaptation to everyday tasks with real-time perception and reasoning.

• Each fingertip has custom, high-fidelity tactile sensors developed in-house, able to detect forces as slight as three grams (delicate enough to handle fragile items like strawberries or paperclips without damage).

• For the home, enhancements include wireless charging, better speakers for voice interaction, and smoother self-docking for recharging.

To reach real-world scale, Figure built BotQ, a robot factory with an initial capacity of 12,000 robots per year and a target of 100,000 by 2029. BotQ uses vertically integrated, in-house manufacturing and automation, with Figure’s own robots building future robots, reducing cost per unit and accelerating iteration while maintaining tight quality control.

Why does it matter?

Figure 03 marks a concrete leap toward general-purpose, everyday humanoid robots. While rivals like Tesla Optimus and Agility Digit remain in prototype stages or niche deployments, Figure is built for scale and mass production, grounded in mature supply chains and in-house component design. Helix’s robust learning and real-time adaptation, combined with home- and factory-ready safety and charging features, means Figure 03 is uniquely positioned to become the first humanoid robot capable of real work in homes, hospitals, and offices.

TRENDING TOOLS

• Clockwise MCP > It brings human-like scheduling powers to AI. Smarter meetings, tasks, and focus time in Claude, Cursor, Relay, and more *

• Opal > Google’s new vibe coding app for creative, intuitive programming

• NeuTTS Air > Clone any voice from just three seconds of audio and generate lifelike speech on your laptop or phone in real time

• Cursor Plan Mode > Cursor's new Plan Mode enables agents to research codebases, draft detailed implementation plans, and let users review or edit them inline before generating code

• Meridian > 50% of searches are now powered by AI (think ChatGPT and Gemini)

• Dragonfly > An AI-powered discovery engine that recommends the best software from a catalog of over 250,000 tools

• FumeDev > Upload a video walkthrough of your app and instantly get a complete Playwright test suite

• DeepMind CodeMender > An autonomous agent powered by Gemini Deep Think models that identifies and patches code vulnerabilities

_{(*) signifies sponsored tool}

THINK PIECES / BRAIN BOOST

• AI is the new social media

• Doomprompting is the new doomscrolling

• Ranking every vibe coding app

• Day 45: Vibe Coding an app until I make $1,000,000

• OpenAI might have just accidentally leaked the top 30 customers who’ve used over 1 trillion tokens

• Ethan Mollick shared a new study asking 2,357 Americans which jobs AI should automate

• More than half of all venture capital money will flow into AI start-ups in 2025

• Real AI agents and real work

• Cognition’s CEO on what comes after code

THE VALLEY GEMS

What’s trending on social today:

THAT’S ALL FOR TODAY

Thank you for reading today’s edition. That’s all for today’s issue.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee

👍️ New reader? Subscribe here

Thanks for being here.

REACH 100K+ READERS

Acquire new customers and drive revenue by partnering with us

Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.

If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.